CN112182959A - Leakage detection method for water distribution pipe network - Google Patents
Leakage detection method for water distribution pipe network Download PDFInfo
- Publication number
- CN112182959A CN112182959A CN202010979018.6A CN202010979018A CN112182959A CN 112182959 A CN112182959 A CN 112182959A CN 202010979018 A CN202010979018 A CN 202010979018A CN 112182959 A CN112182959 A CN 112182959A
- Authority
- CN
- China
- Prior art keywords
- node
- water distribution
- distribution network
- leakage
- probability
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 72
- 238000001514 detection method Methods 0.000 title claims abstract description 23
- 238000000034 method Methods 0.000 claims abstract description 44
- 238000012795 verification Methods 0.000 claims abstract description 32
- 238000012549 training Methods 0.000 claims abstract description 22
- 230000008859 change Effects 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000011835 investigation Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 238000012360 testing method Methods 0.000 abstract description 3
- 238000004458 analytical method Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 239000013505 freshwater Substances 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 230000007774 longterm Effects 0.000 description 2
- 238000000691 measurement method Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000001052 transient effect Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000003321 amplification Effects 0.000 description 1
- 238000012076 audiometry Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000010972 statistical evaluation Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/18—Network design, e.g. design based on topological or interconnect aspects of utility systems, piping, heating ventilation air conditioning [HVAC] or cabling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/08—Fluids
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/14—Pipes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Abstract
The invention discloses a leakage detection method for a water distribution pipe network, which comprises the following steps: the method comprises the following steps: detecting the pressure value p of each node of the water distribution network in a stable operation state; step two: detecting pressure values p of nodes of water distribution network in leakage state0(ii) a Step three: calculating residual error r of each node, wherein r is p-p0Dividing the residual r data of each node into a verification data set and a training data set, and bringing the training data set and the prior data into a Bayes model for trainingTraining to obtain a corrected Bayesian model; verifying by adopting a verification data set, and entering the next step after the verification is passed; step four: and judging the leakage point of the water distribution network according to the probability. The test result shows that the pipe network leakage detection method is less influenced by uncertain factors, has high accuracy and has better practical value.
Description
Technical Field
The invention relates to the technical field of leakage detection of a water distribution pipe network, in particular to a pipe network leakage detection method based on expert knowledge and a Bayesian model.
Background
Water is the most important resource on which human society relies for survival and development. At present, the problems of China in the field of fresh water resources are as follows: on one hand, the fresh water resources are poor, the water environment is seriously polluted, and the treatment is very difficult; on the other hand, the urban water supply system has serious leakage and is not effectively treated for a long time, so that the originally serious shortage condition of fresh water resources becomes snowy and frosted.
All water distribution networks have a certain leakage problem. The leakage detection of the water distribution network is a complex technical work, and the traditional detection methods are equipment methods, including an audio query method, a flow measurement method, a pressure wave amplitude visual positioning method and the like. The audio frequency query method is a commonly used pipe network leak detection method at present, and can be divided into 2 kinds of valve plug audition and ground audition. The technical personnel use the leakage listening rod and the electronic amplification leakage listening instrument to detect the water leakage sound generated by the leakage of the pipeline at the positions of the valve, the fire extinguishing detector, the pipeline exposure position and the like. The method has the advantages of simple equipment, good effect and high cost performance. The defects are that environmental noise has a large influence on audiometry and depends on the experience of workers. Flow measurement refers to the determination of network leaks by electromagnetic, ultrasonic, and other flow meters. The pressure wave amplitude visual positioning method is that the pressure of the pipeline is regulated to change the pressure amplitude of the sound wave around the leakage point, so as to determine the position of the leakage point. These measurement methods all require a large number of technicians to participate in the detection work of the pipe network, and the detection workload is huge because the laying range of the pipe network is wide.
At present, pipe network leakage detection also comprises methods based on models, such as a mass balance method, a leakage detection method based on a transient theory, a data driving method and the like. The mass balance method is based on the principle of conservation of mass, and the core of the method is that when the pipeline does not leak, the inflow mass and the outflow mass of the pipeline should be equal. Although the principle of the method is simple and is easy to implement, the actual underground pipeline is very complex and is not always in a stable state, the method is very sensitive to any disturbance and dynamic change of the pipeline and is easy to generate wrong conclusions, and in addition, the method can only roughly judge a leakage area and cannot accurately position a leakage point. The leakage detection method based on the transient theory means that pressure waves are generated in a pipe network when leakage occurs, the monitoring points can detect the propagation time of the pressure waves reflected from the leakage points, and the leakage points can be positioned by performing frequency domain analysis on the pressure wave signals. The method has the advantages of sensitivity and economy, but the feasibility of the method cannot be verified due to the complexity of the parallel connection of the actual pipe networks at present, and subsequent continuous experimental research is needed. The data driving method refers to a process of detecting leakage by using a data driving model, which comprises the steps of firstly extracting a large amount of hydraulic data which are detected by each monitoring point and are leaked and not leaked, then carrying out statistical analysis and evaluation on the data, then determining regular characteristics of the monitoring point data when the leakage phenomenon of a pipe network occurs according to information obtained by feedback, and when the characteristics of the monitoring point data occur, the leakage phenomenon is suspected. Aiming at a large complex pipe network, the method is the highest accuracy in a model method at present, but the method does not have the hydraulic model as the support data, and the explanation of the hydraulic model is purely dependent on a mathematical statistic model, so that the method may be too much different from the actual situation.
Disclosure of Invention
Aiming at the influence of a plurality of uncertain factors such as complex structure, uncertain node requirements, uncertain leakage amount and leakage position, sensor noise and the like in a water distribution network pipe network, the invention provides a leakage detection method for the water distribution network.
A leakage detection method for a water distribution network comprises the following steps:
the method comprises the following steps: detecting the pressure value p of each node of the water distribution network in a stable operation state (namely, no leakage);
step two: detecting pressure values p of nodes of a distribution network in the event of a leakage condition (i.e. unstable operation condition)0；
Step three: calculating residual error r of each node, wherein r is p-p0Dividing the data of the residual error r of each node into a verification data set and a training data set, and bringing the training data set and the prior data into a Bayesian model for training to obtain a modified Bayesian model;
verifying the corrected Bayes model by adopting a verification data set, if the verification is passed, entering the next step, and if the verification is not passed, bringing the training data set and new prior data into the Bayes model for retraining;
step four: detecting the variation value of total flow of water distribution network, and detecting the pressure value p of each node of water distribution network when the water distribution network is in leakage state0Calculating residual error r of each node with pressure value p of each node of the water distribution network in the stable operation state, wherein r is p-p0And substituting the residual error r of each node into the corrected Bayesian model, calculating to obtain the leakage probability of each node of the water distribution network, and judging the leakage point of the water distribution network according to the probability.
In the first step, the stable operation state refers to a variation value of 0-2% of the total flow of the water distribution pipe network. The change value of the total flow of the water distribution network is maximum flow/minimum flow-100%.
In the second step, the leakage state is a change value of the total flow of the water distribution network of 2.01% to 10%, and the change value of the total flow of the water distribution network is-100% of the maximum flow/the minimum flow.
In step three, the verification data set comprises: 15% -30% of residual r data of each node, wherein the training data set comprises: 70% to 85% of the residual r of each node. Most preferably, the validation data set comprises: 20% of the residual r of each node, said training data set comprising: 80% of the data of the residual r of each node.
The modified Bayesian model is as shown in formula (1):
wherein, P (l)iL r (k)) is a leakage node liWhen the leakage occurs, the residual error is the posterior probability of r (k), which is the basis for judging the leakage position, P (r (k) | li) When leaking the node liProbability of occurrence of residual value r (k), P (l)i) Is a prior probability which is induced by an expert system (i.e. empirically) through inference, wherein P (r (k)) is a normalization factor given by a total probability formula
And (6) obtaining. Wherein, in the calculation of P (r (k)), P (r (k) | li) When leaking the node liProbability of occurrence of residual value r (k), P (l)i) Is a prior probability.
The prior data comprises: the prior probability of each node of the distribution network.
The prior probability is obtained according to experience and is mainly obtained by calculation through experience knowledge, background knowledge and model knowledge:
prior probability of 20% × P1+60％×P2+20％×P3；
Wherein, P1Probability provided for according to empirical knowledge, P2Probability obtained for background investigation, P3Probability obtained for model simulation, P1Is 0.08 to 0.12, P20.1 to 0.13, P30.1 to 0.12.
And verifying the corrected Bayes model, if the verification is passed and the next step is carried out, the verification is not passed, and the training data set and the new prior data are brought into the Bayes model for retraining, specifically comprising:
the verification passing needs to meet the requirement that the verification accuracy rate exceeds 90 percent, and the verification passes;
new prior data is obtained by using the prior probability of 19% × P1+60％×P2+21％×P3And recalculating the prior probability of each node of the water distribution network.
If the corrected Bayes model obtained after retraining fails to be verified again, the prior probability of each node of the water distribution network is 18% multiplied by P1+60％×P2+22％×P3Recalculating, and analogizing until the verification is passed.
Compared with the prior art, the invention has the following advantages:
the leakage detection method for the water distribution network has the advantages of higher accuracy, less calculation amount and high response speed when uncertain factors such as uncertain node requirements, uncertain leakage positions, uncertain leakage amount, sensor noise and the like exist in the network. The EPANET pipe network model established by collecting pipe network design data and hydraulic data is more reliable and can reflect the running condition of the actual pipe network. The pressure difference value of the model during normal operation and leakage is used as data to be judged through the Bayesian model optimized through empirical knowledge, so that the interference of uncertain factors on the model can be reduced, and the detection accuracy of the model on leakage is improved.
The test result shows that the pipe network leakage detection method is less influenced by uncertain factors, has high accuracy and has better practical value.
Drawings
FIG. 1 is a flow chart of a method for detecting leakage in a water distribution network according to the present invention;
FIG. 2 is a schematic diagram of the residual error generated in the method;
FIG. 3 is a diagram of a pipe network established in example one;
FIG. 4 is a flow chart of a method for detecting leakage in a water distribution network according to an embodiment of the present invention.
Detailed Description
As shown in fig. 4, a method for detecting leakage of a water distribution network includes the following steps:
the method comprises the following steps: and detecting the pressure value p of each node of the water distribution network in a stable operation state (namely, no leakage), wherein the stable operation state refers to a change value of the total flow of the water distribution network of 0-2%. The change value of the total flow of the water distribution network is maximum flow/minimum flow-100%.
Step two: detecting pressure values p of nodes of a distribution network in the event of a leakage condition (i.e. unstable operation condition)0The leakage state is 2.01% -10% of the change value of the total flow of the water distribution pipe network, and the change value of the total flow of the water distribution pipe network is the maximum flow/the minimum flow-100%.
Step three: calculating residual error r of each node, wherein r is p-p0Dividing the data of the residual error r of each node into a verification data set and a training data set according to 20% and 80%, and bringing the training data set and the prior data into a Bayes model for training to obtain a modified Bayes model;
the modified Bayesian model is as shown in formula (1):
wherein, P (l)iL r (k)) is a leakage node liWhen the leakage occurs, the residual error is the posterior probability of r (k), which is the basis for judging the leakage position, P (r (k) | li) When leaking the node liProbability of occurrence of residual value r (k), P (l)i) Is a prior probability which is induced by an expert system (i.e. empirically) through inference, wherein P (r (k)) is a normalization factor given by a total probability formula
And (6) obtaining. Wherein, in the calculation of P (r (k)), P (r (k) | li) When leaking the node liProbability of occurrence of residual value r (k), P (l)i) Is a prior probability.
The prior data comprises: the prior probability of each node of the distribution network.
The prior probability is obtained according to experience and is mainly obtained by calculation through experience knowledge, background knowledge and model knowledge:
prior probability of 20% × P1+60％×P2+20％×P3；
Wherein, P1Probability provided for according to empirical knowledge, P2Probability obtained for background investigation, P3Probability obtained for model simulation, P1Is 0.08 to 0.12, P20.1 to 0.13, P30.1 to 0.12.
Verifying the corrected Bayes model by adopting a verification data set, if the verification is passed, entering the next step, and if the verification is not passed, bringing the training data set and new prior data into the Bayes model for retraining;
the verification passing needs to meet the requirement that the verification accuracy rate exceeds 90 percent, and the verification passes;
new prior data is obtained by using the prior probability of 19% × P1+60％×P2+21％×P3And recalculating the prior probability of each node of the water distribution network.
If the corrected Bayes model obtained after retraining fails to be verified again, the prior probability of each node of the water distribution network is 18% multiplied by P1+60％×P2+22％×P3Recalculated, analogize in turn (i.e. P1Decrease of coefficient of (A), P3Increases) until the verification is passed.
Step four: detecting the variation value of total flow of water distribution network, and detecting the pressure value p of each node of water distribution network when the water distribution network is in leakage state0Calculating residual error r of each node with pressure value p of each node of the water distribution network in the stable operation state, wherein r is p-p0And substituting the residual error r of each node into the corrected Bayesian model, calculating to obtain the leakage probability of each node of the water distribution network, and judging the leakage point of the water distribution network according to the probability.
As shown in fig. 1, a method for detecting leakage of a water distribution network includes the following steps:
the method comprises the following steps: collecting related data such as pipe network design data and hydraulic data, and establishing a pipe network model through EPANET software;
by collecting background data such as a pipe network layout diagram, hydraulic data and the like, a pipe network model is established by utilizing EPANET software, and the pipe network model can better reflect the actual pipe network operation condition;
step two: and (4) reasoning and generating the prior knowledge required by the Bayesian model through expert knowledge. Each piece of leakage knowledge in the expert knowledge base comprises leakage nodes, leakage performance parameters and the like.
The acquisition of expert knowledge mainly comprises three parts;
(1) and (4) knowledge of experience. Empirical knowledge refers to heuristic knowledge accumulated by experts in the field related to leakage of a pipe network in long-term fault detection through practice about how to hypothesize, judge and identify leakage according to observed symptoms.
(2) Background knowledge. Background knowledge refers to knowledge of design, installation, maintenance, etc. related to the pipe network, long-term operating conditions, locations where leaks are likely to occur, and corresponding processing results, etc.
(3) And (4) knowledge of the model. The model knowledge has more determinacy and objectivity, the occurrence process of the pipe network leakage is quantitatively analyzed through the established EPANET pipe network model, and the unexpected leakage can be detected based on the leakage analysis of the pipe network model;
deducing nodes which are easy to leak in the pipe network system by using established facts and rules stored in an expert knowledge base, and converting a conclusion obtained by the deduction into probability distribution to be used as prior knowledge of a Bayesian model;
step three: as shown in fig. 1 and 2, leakage simulation is performed on the EPANET pipe network model established in the first step, a difference value between normal operation (no leakage) and leakage at each node is generated, knowledge generated by inference of the expert system in the second step is used as prior knowledge, and a bayesian model is established by combining the prior knowledge and the prior knowledge;
simulating the leakage condition of each node through an EPANET pipe network model, generating residual errors of each node during normal operation and leakage, and calculating the residual errors r of each node, wherein r is p-p0。
Assuming that N nodes in a pipe network are all likely to generate leakage with the value of X% -Y% of total flow, and the node with the leakage is recorded as liN, M sensors in the network measure the node pressureIs denoted by pjJ is 1. Defining an actual residual error as a difference value between a measured value provided by the sensor and a model predicted value, wherein in each moment k, the residual error correspondingly generated in the model is r (k), and training and verifying the Bayesian model by utilizing residual error data.
All residuals will be activated to some extent when leakage actually occurs. Each time the residual is sampled, bayesian theorem is applied:
wherein, P (l)iL r (k)) is the leak liWhen the leakage occurs, the residual error is the posterior probability of r (k), which is the basis for judging the leakage position, P (r (k) | li) When leakage is causediProbability of occurrence of residual value r (k), P (l)i) Is a prior probability obtained by inference induction of an expert system, wherein P (r (k)) is a normalization factor given according to a total probability formula
And (6) obtaining.
Step four: and performing inference analysis on the online data of the pipe network according to the corrected Bayesian model, taking the change value 2% of the total flow of the water distribution pipe network as a threshold, detecting any residual exceeding the threshold when the residual is generated to obtain the probability of possible leakage of each node in the pipe network, outputting and displaying, wherein the node represented by the maximum probability value is the most likely position of leakage, and giving a corresponding alarm signal.
The method for diagnosing pipe network leakage according to the present invention will be described in detail with reference to fig. 3. The examples of the invention are as follows:
firstly, an EPANET circular pipe network model is established according to related pipe network design data and water power. In the pipe network model, 1 is a reservoir and provides a total water head for the whole pipe network. The water is boosted between the reservoir and the node 1 through the pump station, and each node is guaranteed to have enough water pressure. The basin 1 receives water that flows out of the pipe network. Nodes 1-9 are demand points, each having a different demand. Some of the node parameters are shown in table 1.
Table 1 partial node parameters
| Node sequence number | Basic Water demand/LPS | Elevation/m |
| 1 | 14.55 | 20 |
| 2 | 51.17 | 18 |
| 3 | 27.65 | 15 |
| 4 | 16.15 | 20 |
| 5 | 30.70 | 20 |
The pipe diameters of the pipelines 1-9 are between 100 and 400 mm. Some pipeline parameters are shown in table 2.
TABLE 2 partial pipeline parameters
| Number of pipes | Length/m | Diameter/mm | Flow velocity/m.s at steady state-1 |
| 3 | 650 | 400 | 3.10 |
| 4 | 550 | 400 | 2.22 |
| 5 | 165 | 300 | 1.43 |
| 6 | 350 | 300 | 0.85 |
| 7 | 180 | 300 | 1.02 |
Assume that 2% -5% leakage occurs at nodes 1-9. Taking the leakage at the node 1 as an example, the aperture of the leakage point is 2-30mm, the leakage amount is equal-difference simulated for 100 times between 2% and 5%, and the pressure value at the node 1-9 is recorded when the leakage amount changes every time. The remaining nodes are sequentially subjected to the above operation, and 900 groups of data are obtained by sampling in total. For example, the leakage occurs at node 1, and part of the data is shown in table 3.
TABLE 3 pressure at each node in case of leakage at node 1
| Node leak aperture | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |
| 5.12 | 58.73 | 37.97 | 28.54 | 56.98 | 36.54 | 28.61 | 56.47 | 30.12 | 25.29 |
| 5.38 | 58.73 | 37.96 | 28.54 | 56.98 | 36.54 | 28.60 | 56.46 | 30.12 | 25.29 |
| 5.64 | 58.72 | 37.96 | 28.54 | 56.97 | 36.53 | 28.60 | 56.46 | 30.11 | 25.29 |
| 5.90 | 58.72 | 37.96 | 28.53 | 56.97 | 36.53 | 28.60 | 56.45 | 30.11 | 25.28 |
| 6.16 | 58.71 | 37.95 | 28.53 | 56.97 | 36.52 | 28.59 | 56.45 | 30.11 | 25.28 |
| 6.42 | 58.71 | 37.95 | 28.53 | 56.96 | 36.52 | 28.59 | 56.44 | 30.10 | 25.28 |
| 6.68 | 58.70 | 37.94 | 28.52 | 56.95 | 36.52 | 28.59 | 56.44 | 30.10 | 25.27 |
| 6.94 | 58.70 | 37.94 | 28.52 | 56.95 | 36.51 | 28.58 | 56.43 | 30.09 | 25.27 |
| 7.20 | 58.69 | 37.93 | 28.51 | 56.94 | 36.51 | 28.58 | 56.43 | 30.09 | 25.26 |
| 7.46 | 58.69 | 37.93 | 28.51 | 56.94 | 36.50 | 28.57 | 56.42 | 30.08 | 25.26 |
| 7.72 | 58.68 | 37.92 | 28.51 | 56.93 | 36.49 | 28.57 | 56.41 | 30.08 | 25.25 |
| 7.98 | 58.67 | 37.92 | 28.50 | 56.92 | 36.49 | 28.56 | 56.41 | 30.07 | 25.25 |
| 8.24 | 58.66 | 37.91 | 28.49 | 56.91 | 36.48 | 28.56 | 56.40 | 30.06 | 25.24 |
| 8.50 | 58.66 | 37.90 | 28.49 | 56.91 | 36.48 | 28.55 | 56.39 | 30.06 | 25.23 |
| 8.76 | 58.65 | 37.90 | 28.48 | 56.90 | 36.47 | 28.55 | 56.38 | 30.05 | 25.23 |
The EPANET pipe network model is evaluated through expert knowledge, probability distribution of each node is obtained and used as prior knowledge of the Bayesian model, and specific prior probabilities are shown in a table 4.
TABLE 4 priori probabilities derived from expert knowledge
| Node point | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |
| P1 | 0.120 | 0.095 | 0.140 | 0.138 | 0.110 | 0.110 | 0.112 | 0.093 | 0.082 |
| P2 | 0.130 | 0.110 | 0.100 | 0.100 | 0.110 | 0.110 | 0.120 | 0.115 | 0.105 |
| P3 | 0.115 | 0.100 | 0.115 | 0.107 | 0.115 | 0.115 | 0.118 | 0.112 | 0.103 |
| Prior probability | 0.125 | 0.105 | 0.111 | 0.109 | 0.111 | 0.111 | 0.118 | 0.110 | 0.100 |
Wherein, P1To provide probabilities based on empirical knowledge, P2Probability obtained for background investigation, P3The resulting probabilities are simulated for the model.
And establishing a Bayesian model basic frame through Python, and randomly segmenting total data, wherein 80% of the total data is used as a training data set to train the rest parameters of the Bayesian model, and 20% of the total data is used as a verification data set to verify the accuracy of the Bayesian model.
The established Bayesian model is applied to an actual pipe network, online data of the pipe network are subjected to reasoning analysis, 1% of total flow in the pipe network is used as a threshold, when a residual error is generated, the probability that each node in the pipe network is likely to leak is obtained by detecting that a certain residual error exceeds the threshold, the probability is output and displayed, and a corresponding alarm signal is given.
Taking the leakage occurring at the node 1 as an example, the leakage amount exceeds 1% of the total flow at this time, performing leakage detection, and subtracting the pressure value of each node when no leakage occurs at the node 1 from the pressure value of each node when leakage occurs to obtain the change characteristic of the residual error when leakage occurs at the node 1. The leakage residuals at each node are shown in table 5.
TABLE 5 residual error at each point with 7.2mm leak aperture at node 1
And (3) introducing the pressure residual errors on the table into a Bayes model, and applying a Bayes formula:
the posterior probabilities of the respective nodes can be obtained as shown in table 6.
TABLE 6A posteriori probabilities for each node
By comparing the posterior probability of each node, the posterior probability of the node 1 is the maximum under each leakage aperture, so that the leakage is presumed to occur at the node 1, and a corresponding alarm signal is sent out.
The accuracy rate of the Bayesian model is 94.5%, compared with the accuracy rate of the Bayesian model without expert knowledge under the same training set and verification set which is 85.8%, the accuracy rate is obviously improved, and the improvement of the accuracy rate of the Bayesian model by the expert knowledge and the feasibility of the method in practical application are proved. The test result shows that the pipe network leakage detection method is less influenced by uncertain factors, has high accuracy and has better practical value.
Claims (10)
1. A water distribution network leakage detection method is characterized by comprising the following steps:
the method comprises the following steps: detecting the pressure value p of each node of the water distribution network in a stable operation state;
step two: detecting pressure values p of nodes of water distribution network in leakage state0；
Step three: calculating residual error r of each node, wherein r is p-p0Dividing the residual r data of each node into a verification data set and a training data set, and collecting the training data set and the training data setThe prior data is brought into a Bayesian model for training to obtain a corrected Bayesian model;
verifying the corrected Bayes model by adopting a verification data set, if the verification is passed, entering the next step, and if the verification is not passed, bringing the training data set and new prior data into the Bayes model for retraining;
step four: detecting the variation value of total flow of water distribution network, and detecting the pressure value p of each node of water distribution network when the water distribution network is in leakage state0Calculating residual error r of each node with pressure value p of each node of the water distribution network in the stable operation state, wherein r is p-p0And substituting the residual error r of each node into the corrected Bayesian model, calculating to obtain the leakage probability of each node of the water distribution network, and judging the leakage point of the water distribution network according to the probability.
2. The method according to claim 1, wherein in the first step, the steady operation state is a variation value of the total flow rate of the water distribution network of 0 to 2%, and the variation value of the total flow rate of the water distribution network is-100% of the maximum flow rate/the minimum flow rate.
3. The method according to claim 1, wherein in the second step, the leakage state is a value of 2.01% to 10% of a change in the total flow rate of the water distribution network, and the value of the change in the total flow rate of the water distribution network is-100% of a maximum flow rate/a minimum flow rate.
4. The method according to claim 1, wherein the verification data set comprises: 15% -30% of residual r data of each node, wherein the training data set comprises: 70% to 85% of the residual r of each node.
5. The method for detecting leakage in a water distribution network according to claim 1, wherein in step three, the modified bayesian model is represented by equation (1):
wherein, P (l)iL r (k)) is the leakage node l at time kiThe posterior probability of the generated residual error r (k) is the basis of the judgment of the leakage position, and P (r (k) | li) When leaking the node liProbability of occurrence of residual value r (k), P (l)i) For prior probability, where P (r (k)) is a normalization factor given according to the total probability formula, which is
Obtaining P (r (k) | l in the calculation of P (r (k))i) When leaking the node liProbability of occurrence of residual value r (k), P (l)i) Is a prior probability.
6. The method according to claim 1, wherein the prior data comprises: the prior probability of each node of the distribution network.
7. The method according to claim 1, wherein in step three, the prior probability is 20% × P1+60％×P2+20％×P3Wherein P is1Probability provided for according to empirical knowledge, P2Probability obtained for background investigation, P3The resulting probabilities are simulated for the model.
8. The method for detecting leakage in water distribution network as claimed in claim 7, wherein P is the number P in the third step1Is 0.08 to 0.12, P20.1 to 0.13, P30.1 to 0.12.
9. The method for detecting leakage in a water distribution network as claimed in claim 7, wherein in step three, the verification pass is required to satisfy the verification accuracy rate exceeding 90%.
10. The method according to claim 1, wherein in step three, the new prior data is obtained by using a prior probability of 19% × P1+60％×P2+21％×P3Recalculating the prior probability of each node of the water distribution network;
if the corrected Bayes model obtained after retraining fails to be verified again, the prior probability of each node of the water distribution network is 18% multiplied by P1+60％×P2+22％×P3Recalculating, and analogizing until the verification is passed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010979018.6A CN112182959A (en) | 2020-09-17 | 2020-09-17 | Leakage detection method for water distribution pipe network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010979018.6A CN112182959A (en) | 2020-09-17 | 2020-09-17 | Leakage detection method for water distribution pipe network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112182959A true CN112182959A (en) | 2021-01-05 |
Family
ID=73920020
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010979018.6A Pending CN112182959A (en) | 2020-09-17 | 2020-09-17 | Leakage detection method for water distribution pipe network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112182959A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113970073A (en) * | 2021-11-11 | 2022-01-25 | 吉林大学 | ResNet-based water supply pipe network leakage accurate positioning method |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111412391A (en) * | 2019-01-04 | 2020-07-14 | 合肥暖流信息科技有限公司 | Pipe network leakage detection method and system |
-
2020
- 2020-09-17 CN CN202010979018.6A patent/CN112182959A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111412391A (en) * | 2019-01-04 | 2020-07-14 | 合肥暖流信息科技有限公司 | Pipe network leakage detection method and system |
Non-Patent Citations (1)
| Title |
|---|
| 李霞等: "基于贝叶斯理论的城市供水管网泄漏在线检测与定位", 《给水排水》, vol. 32, no. 12, 10 December 2006 (2006-12-10), pages 97 - 99 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113970073A (en) * | 2021-11-11 | 2022-01-25 | 吉林大学 | ResNet-based water supply pipe network leakage accurate positioning method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR100426227B1 (en) | Methods, devices and systems for identifying subjects | |
| Song et al. | Gas leak detection in galvanised steel pipe with internal flow noise using convolutional neural network | |
| US20160011072A1 (en) | Systems and methods for determining a leak rate through an opening using acoustical sensors | |
| CN112097126B (en) | Water supply network pipe burst pipeline accurate identification method based on deep neural network | |
| CN106844814A (en) | A kind of large complicated gas distributing system system leak detection method | |
| Wang et al. | Experimental study on water pipeline leak using In-Pipe acoustic signal analysis and artificial neural network prediction | |
| CN101592288A (en) | A kind of method for identifying pipeline leakage | |
| US20220011275A1 (en) | Method and system to analyse pipeline condition | |
| CN111695465B (en) | Pipe network fault diagnosis and positioning method and system based on pressure wave mode identification | |
| US8010468B2 (en) | Method for wafer analysis with artificial neural network and system thereof | |
| CN104535281A (en) | Airport apron pipe network impermeability testing device and method based on pressure correlation method | |
| KR20210064619A (en) | Apparatus for detecting abnormal situation using deep learning ensemble model and method thereof | |
| Li et al. | Leakage detection of HVAC pipeline network based on pressure signal diagnosis | |
| CN109344784A (en) | A kind of leakage orienting method merging hydrophone and Wavelet Denoising Method | |
| CN112182959A (en) | Leakage detection method for water distribution pipe network | |
| CN105674065A (en) | Variational mode decomposition-based method for locating leakage point of pipeline by acoustic emission | |
| Okeya et al. | Locating pipe bursts in a district metered area via online hydraulic modelling | |
| CN112963740B (en) | Method for monitoring and positioning leakage of fire fighting pipeline of convertor station | |
| WO2019056121A1 (en) | Methods for detecting pipeline weakening | |
| Wu et al. | Novel signal denoising approach for acoustic leak detection | |
| Ribeiro et al. | Leak detection modeling of a pipeline using echo state neural networks | |
| CN112097125B (en) | Water supply pipe network pipe burst detection and positioning method based on self-adaptive checking | |
| Wu et al. | Pressure-dependent leakage detection method compared with conventional techniques | |
| JPH10221197A (en) | Method and device for detecting leakage | |
| Salam et al. | Application of SVM and ELM methods to predict location and magnitude leakage of pipelines on water distribution network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |